| Camera-based gaze estimation technology has become an active field in Human Computer Interaction (HCI) which has been widely used in various fields. Currently, most gaze estimation technology base on the Pupil-Cornea Reflection Technique need to use extra auxiliary light source, and user should make his head as still as possible. Those two points limit the popularization of the gaze estimation.To avoid using extra light sources and improve the tolerance of head movements, this dissertation proposes a 3D model gaze estimation method based on a single camera. This method is established on the accurate eye model, which based on the spatial relationships of the eye and the screen, set up a 3D model of gaze estimation allows the user at any distance and under natural head movement. In this paper,3D model gaze estimation method requires the user’s head calibration, then derive the equations to solve the corner point, the eyeball center and the pupil center, and the user’s gaze direction, then we can get the point of gaze. In addition, this article joins the system calibration, and it is used to improve the accuracy of the gaze estimation.This paper designed a series of experiments to get the distance between user and the screen, and the point of gaze estimation accuracy under different conditions. During the experiment, user is required under different distance from the screen, look at several predefined points on the screen, each gaze state is captured by a camera, and then extract the facial feature points, finally estimate the distance and the point of gaze.The experimental results show that the accuracy of this algorithm is less than 5°and have a good stability. The head rest of the gaze estimation has better accuracy than the situation of head movement. The system can be made for users of different constraints based on different accuracy requirements. At last, future research base on gaze estimation model is suggested by the guide concluded from the dissertation to promote accuracy. |
PDF Full Text Request